Metraipolis › archive › dc_102770d98eb0

Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal — arXiv:2609.04482

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.04482
License (as archived)
CC0-1.0
Archived
2026-10-16T22:35:07.421Z
SHA-256 of the archived text
f493f0ffd752593d4fe3fca3df0562bb7c81bda10960da8ba426ab1b6e39d24c
Stable URL of this page
https://metraipolis.com/documento/dc_102770d98eb0

How to cite

Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal — arXiv:2609.04482 — arXiv (en), https://arxiv.org/abs/2609.04482 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_102770d98eb0 (Metraipolis archive dc_102770d98eb0, sha256 f493f0ffd752593d4fe3fca3df0562bb7c81bda10960da8ba426ab1b6e39d24c)
BibTeX
@misc{dc102770d98eb0,
  title = {Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal — arXiv:2609.04482},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.04482},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_102770d98eb0; sha256 f493f0ffd752593d4fe3fca3df0562bb7c81bda10960da8ba426ab1b6e39d24c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04482. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Safety alignment is usually posed as a topic-level question: is this subject harmful? Deployments ask a narrower one. A civics tutor and a public-sector assistant may share a base model yet need different boundaries inside the same topic, refusing targeted political manipulation while still answering factual questions about the same election. We formulate this as narrow-boundary safety and introduce an offline self-generated framework combining controlled topic generation, coverage repair, in-distribution compensation data, and harmful-benign pairs for training and evaluation. Single-shot generation leaves 19.88% of prompts without accepted refusal traces, whereas escalating retries leave 0.20%. On political persuasion with Qwen3-8B, training on refusal data completed through Escalate increases target-domain refusal from 9.47% to 84.75% and reduces the mean unsafe-response rate across three broader harmfulness benchmarks from 26.26% to 0.14%, but increases XSTest over-refusal from 2.00% to 74.00%. In a separate matched comparison, replacing external responses with verified target-model responses reduces over-refusal from 15.20% to 5.20%. Boundary-pair data reduces comply-side over-refusal on held-out pairs from 32.94% to 4.16%, while harmful-side refusal decreases only from 91.88% to 87.72%. These results show that data composition controls the safety and usability trade-off, and that safety alignment should be evaluated on both sides of the intended refusal boundary.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-07 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.04482 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-16T22:21:16.266Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).